Model import crashes the Unity Editor

When importing a glTF model in Unity using glTFast and the KtxUnity package the Unity editor crashes. This is observed when KtxUnity is added via the file system (either the current main branch or the commit tagged 2.0.1. If I add the package via git URL glTFast claims that the KtxUnity package is not even available (see the issue in the glTFast repo https://github.com/atteneder/glTFast/issues/381).

Expected behavior

The model is imported correctly. Addtitionally, a model import should never crash the editor or a runtime application, even for completely broken models.

OS: macOS Unity editor version: 2021.2.19f1 KtxUnity version: main / 2.0.1 Test model: https://github.com/KhronosGroup/glTF/files/4132437/Avocado.zip

package.json

    "com.atteneder.gltfast": "4.6.0",
    "com.atteneder.ktx": "file:../mypackages/KtxUnity",

editor.log

Start importing Assets/Avocado/Avocado.gltf using Guid(a7299cc32fc4d43a3856f49f0bc3efcf) Importer(-1,00000000000000000000000000000000) Assertion failed: (colorModel == KHR_DF_MODEL_UASTC), function ktxTexture2_TranscodeBasis, file /Users/runner/work/KTX-Software-Unity/KTX-Software-Unity/KTX-Software/lib/basis_transcode.cpp, line 177.
Obtained 0 stack frames.
Launching bug reporter
Assertion failed: (colorModel == KHR_DF_MODEL_UASTC), function ktxTexture2_TranscodeBasis, file /Users/runner/work/KTX-Software-Unity/KTX-Software-Unity/KTX-Software/lib/basis_transcode.cpp, line 177.
Obtained 0 stack frames.
Launching bug reporter
Assertion failed: (colorModel == KHR_DF_MODEL_UASTC), function ktxTexture2_TranscodeBasis, file /Users/runner/work/KTX-Software-Unity/KTX-Software-Unity/KTX-Software/lib/basis_transcode.cpp, line 177.
Obtained 0 stack frames.

[atteneder]

@MarkCallow That’s a great question! I assumed asserts are active in release builds as well, but that’s not true. Turns out my macOS build was misconfigured. After re-building the lib, libktx returns a proper error and does not crash anymore.

Thanks for pointing it out!

[atteneder]

@seyfeb You’re right about crashes are not a reasonable reaction to corrupt data. Since the assertion is inside the third party code, I forwarded the request in a discussion. Let’s see what their answer is.